Normalize f32 helper tails for ggml vec ops

Author: Aaron

ggml/src/ggml-cpu/vec.h

@@ -78,72 +78,60 @@ inline static void ggml_vec_add_f16 (const int n, ggml_fp16_t * z, const ggml_fp
     }
 }
 inline static void ggml_vec_add1_f32(const int n, float * z, const float * x, const float v) {
+    int i = 0;
 #if defined(GGML_SIMD)
     const int np = (n & ~(GGML_F32_STEP - 1));
 
     GGML_F32_VEC vv = GGML_F32_VEC_SET1(v);
 
-    for (int i = 0; i < np; i += GGML_F32_STEP) {
+    for (; i < np; i += GGML_F32_STEP) {
         for (int j = 0; j < GGML_F32_ARR; ++j) {
             GGML_F32_VEC ax = GGML_F32_VEC_LOAD(x + i + j*GGML_F32_EPR);
             GGML_F32_VEC az = GGML_F32_VEC_ADD(ax, vv);
             GGML_F32_VEC_STORE(z + i + j*GGML_F32_EPR, az);
         }
     }
-
-    for (int i = np; i < n; ++i) {
-        z[i] = x[i] + v;
-    }
-#else
-    for (int i = 0; i < n; ++i) {
+#endif
+    for (; i < n; ++i) {
         z[i] = x[i] + v;
     }
-#endif
 }
 inline static void ggml_vec_acc_f32 (const int n, float * y, const float * x) {
+    int i = 0;
 #if defined(GGML_SIMD)
     const int np = (n & ~(GGML_F32_STEP - 1));
 
-    for (int i = 0; i < np; i += GGML_F32_STEP) {
+    for (; i < np; i += GGML_F32_STEP) {
         for (int j = 0; j < GGML_F32_ARR; ++j) {
             GGML_F32_VEC ay = GGML_F32_VEC_LOAD(y + i + j*GGML_F32_EPR);
             GGML_F32_VEC ax = GGML_F32_VEC_LOAD(x + i + j*GGML_F32_EPR);
             ay = GGML_F32_VEC_ADD(ay, ax);
             GGML_F32_VEC_STORE(y + i + j*GGML_F32_EPR, ay);
         }
     }
-
-    for (int i = np; i < n; ++i) {
-        y[i] += x[i];
-    }
-#else
-    for (int i = 0; i < n; ++i) {
+#endif
+    for (; i < n; ++i) {
         y[i] += x[i];
     }
-#endif
 }
 inline static void ggml_vec_acc1_f32(const int n, float * y, const float v) {
+    int i = 0;
 #if defined(GGML_SIMD)
     const int np = (n & ~(GGML_F32_STEP - 1));
 
     GGML_F32_VEC vv = GGML_F32_VEC_SET1(v);
 
-    for (int i = 0; i < np; i += GGML_F32_STEP) {
+    for (; i < np; i += GGML_F32_STEP) {
         for (int j = 0; j < GGML_F32_ARR; ++j) {
             GGML_F32_VEC ay = GGML_F32_VEC_LOAD(y + i + j*GGML_F32_EPR);
             ay = GGML_F32_VEC_ADD(ay, vv);
             GGML_F32_VEC_STORE(y + i + j*GGML_F32_EPR, ay);
         }
     }
-
-    for (int i = np; i < n; ++i) {
-        y[i] += v;
-    }
-#else
-    for (int i = 0; i < n; ++i) {
+#endif
+    for (; i < n; ++i) {
         y[i] += v;
     }
-#endif
 }
 inline static void ggml_vec_sub_f32 (const int n, float * z, const float * x, const float * y) { for (int i = 0; i < n; ++i) z[i]  = x[i] - y[i]; }
 inline static void ggml_vec_sub_f16 (const int n, ggml_fp16_t * z, const ggml_fp16_t * x, const ggml_fp16_t * y) {
@@ -152,25 +140,21 @@ inline static void ggml_vec_sub_f16 (const int n, ggml_fp16_t * z, const ggml_fp
     }
 }
 inline static void ggml_vec_set_f32 (const int n, float * x, const float v) {
+    int i = 0;
 #if defined(GGML_SIMD)
     const int np = (n & ~(GGML_F32_STEP - 1));
 
     GGML_F32_VEC vx = GGML_F32_VEC_SET1(v);
 
-    for (int i = 0; i < np; i += GGML_F32_STEP) {
+    for (; i < np; i += GGML_F32_STEP) {
         for (int j = 0; j < GGML_F32_ARR; ++j) {
             GGML_F32_VEC_STORE(x + i + j*GGML_F32_EPR, vx);
         }
     }
-
-    for (int i = np; i < n; ++i) {
-        x[i] = v;
-    }
-#else
-    for (int i = 0; i < n; ++i) {
+#endif
+    for (; i < n; ++i) {
         x[i] = v;
     }
-#endif
 }
 inline static void ggml_vec_cpy_f32 (const int n, float * y, const float * x)                  { for (int i = 0; i < n; ++i) y[i]  = x[i];        }
 inline static void ggml_vec_neg_f32 (const int n, float * y, const float * x)                  { for (int i = 0; i < n; ++i) y[i]  = -x[i];       }
@@ -181,26 +165,22 @@ inline static void ggml_vec_neg_f16 (const int n, ggml_fp16_t * y, const ggml_fp
 }
 
 inline static void ggml_vec_mul_f32 (const int n, float * z, const float * x, const float * y) {
+    int i = 0;
 #if defined(GGML_SIMD)
     const int np = (n & ~(GGML_F32_STEP - 1));
 
-    for (int i = 0; i < np; i += GGML_F32_STEP) {
+    for (; i < np; i += GGML_F32_STEP) {
         for (int j = 0; j < GGML_F32_ARR; ++j) {
             GGML_F32_VEC ax = GGML_F32_VEC_LOAD(x + i + j*GGML_F32_EPR);
             GGML_F32_VEC ay = GGML_F32_VEC_LOAD(y + i + j*GGML_F32_EPR);
             GGML_F32_VEC az = GGML_F32_VEC_MUL(ax, ay);
             GGML_F32_VEC_STORE(z + i + j*GGML_F32_EPR, az);
         }
     }
-
-    for (int i = np; i < n; ++i) {
-        z[i] = x[i]*y[i];
-    }
-#else
-    for (int i = 0; i < n; ++i) {
+#endif
+    for (; i < n; ++i) {
         z[i] = x[i]*y[i];
     }
-#endif
 }
 inline static void ggml_vec_mul_f16 (const int n, ggml_fp16_t * z, const ggml_fp16_t * x, const ggml_fp16_t * y) {
     for (int i = 0; i < n; ++i) {